Phosphorus-31 NMR study of the interaction of inorganic phosphate

Superoxide Dismutase. Duarte Mota de Freitas,* 1**1 Claudio Luchinat,1 Lucia Band,1 Ivano Bertini,*1 and Joan Selverstone Valentine*8. Received Octobe...
0 downloads 0 Views 512KB Size
Inorg. Chem. 1987, 26, 2788-2791

2788

Contribution from the Departments of Chemistry, Loyola University of Chicago, Chicago, Illinois 60626, and University of Florence, Florence, Italy, and Department of Chemistry and Biochemistry and Molecular Biology Institute, University of California, Los Angeles, California 90024

31PNMR Study of the Interaction of Inorganic Phosphate with Bovine Copper-Zinc Superoxide Dismutase Duarte Mota de Freitas,? Claudio Luchinat,t Lucia Banci,t Ivano Bertini,*t and Joan Selverstone Valentine*$ Received October 15, I986 Paramagnetic effects on IlP phosphate resonances caused by Cu(I1) ions in native and phenylglyoxal-modified bovine Cu,Znsuperoxide dismutase have been used to monitor the interaction of phosphate with these proteins. T2values are found to be 70 times smaller than T,, indicating that some mechanisms, as yet undefined, contribute to the line width. Using TI measurements, we have determined that the affinity constants for phosphate binding to the native protein are 20 f 4 and 34 f 3 M-I at pH 8.0 and 7.0, respectively, and that the Cu(I1)-phosphate distance is 5.3 8,. At pH 6.3, two binding sites are observed, one at a distance > 7 8, with an affinity constant > l o 0 M-l and another at approximately 5 8, with an affinity constant of 10 M-I. Modification of the protein with phenylglyoxal causes the affinity of phosphate for the same sites to decrease by a factor of 3 at pH 6.3. These results indicate that phosphate does not bind directly to Cu(I1) but to a site close by. We conclude that the site of phosphate binding is Arg-141, which is known from X-ray structural evidence to be located approximately 5 A from the copper center.

Introduction Bovine erythrocyte Cu,Zn-superoxidedismutase (Cu,Zn-SOD) [EC1.15.1.1] is a metalloenzyme of mol wt 31 200 that is comprised of two identical subunits, each of which contains one copper and one zinc ion in close proximity. I t is a particularly well characterized metalloenzyme system' in that its amino acid sequence and X-ray crystal structure are known, the latter t o 2-A resol~tion.~ This ~ ~ enzyme is found in t h e cytosol of almost all eukaryotic cells and in some bacteria and has been proposed t o function in the cell as a protective a g e n t against the effects of s~peroxide.~ X-ray crystal structural studies of bovine Cu,Zn-SOD indicate t h a t the positively charged side chains of Arg-141, Lys-120, and Lys- 134 are located in the vicinity of the active site, 5, 12, and 13 A, respectively, away from t h e copper I t is believed that these residues are present in predominantly hydrophobic portions of t h e channel t h a t provides access to t h e active site and that they provide electrostatic guidance to the substrate, superoxide, and other anion^.^,^ The site of reactivity for superoxide with the oxidized form of t h e enzyme is t h e copper(I1) ion, which is also t h e binding site for several anions, as indicated by pronounced spectral shifts observed upon anion binding.' Inhibitory effects of these anions have been attributed t o this type of binding.' Phosphate has also been shown t o interact with bovine Cu,Zn-SOD,'-lO but no visible or ESR spectral changes were observed in solutions of Cu,Zn-SOD even a t high concentrations of phosphate." In an earlier paper," it was reported that the SOD activity of native bovine Cu,Zn-SOD, as well as its affinity for cyanide and azide, measured a t constant ionic strength, decreased significantly in t h e presence of increasing concentrations of phosphate. When these experiments were repeated with Cu,Zn-SOD t h a t had been chemically modified a t Arg-141 with phenylglyoxal, by contrast, little dependence of either SOD activity or anion binding on phosphate concentration was observed. These results were interpreted as indicating t h a t the inhibitory effect of phosphate on bovine Cu,Zn-SOD was due primarily t o binding of phosphate t o t h e side chain of Arg- 14 1 with consequent neutralization of its positive charge. A preliminary study using 31PNMR was also reported in which line broadening of phosphate resonances of inorganic phosphate and adenine nucleotides caused by t h e paramagnetic Cu(I1) center w a s described.I2-l4 In this paper, we present a quantitative analysis of t h e binding of inorganic phosphate t o bovine C u , Z n - S O D . P a r a m a g n e t i c effects on 31Pphosphate resonances caused b y copper(I1) ions present a t t h e active site of native and chemically modified proteins Loyola University. 'University of Florence. UCLA.

+

0020-1669/87/1326-2788$01.50/0

have been used to monitor the interaction of phosphate with these proteins, and 31PTI measurements have allowed us to discriminate between direct metal-to-phosphate binding and binding t o other locations within 10

Experimental Section Bovine liver Cu,Zn-SOD was purchased as a lyophilized powder from Diagnostic Data, Inc. (Mountain View, CA). Phenylglyoxal, sodium azide, HEPES [4-(2-hydroxyethyl)- 1-piperazineethanesulfonicacid], and sodium dithionite were purchased from Sigma, potassium phosphate dibasic trihydrate was purchased from Mallinckrodt, Inc., and deuterium oxide (99.8%) was from Cambridge Isotope Laboratories. All were used as received. Bovine Cu,Zn-SOD was chemically modified at Arg-141 with phenylglyoxal by the method of Malinowski and Fridovich.I6 Protein concentrations were determined by using the absorbances at 260 nm ( t = 10300 M-I sd) and 680 nm (e = 300 M-' s-l).l The two values always agreed within 5%. Native and Arg-modified SOD were reduced by addition of sufficient sodium dithionite to bleach the color of the protein. IIP N M R measurements were made either at 32.4 MHz on a Varian FT80 or at 36.4 MHz on a Bruker CXP90 spectrometer at the ambient probe temperature (-30 "C). Some of the measurements were repeated at 121.3 MHz on a Bruker CXP300 N M R spectrometer. Longitudinal

Valentine, J. S.; Pantoliano, M. W. In Copper Proteins; Spiro, T. G., Ed.; Wiley: New York, 1981; pp 291-358. Tainer, J. A,; Getzoff, E. D.; Beem, K. M.; Richardson, J. S.; Richardson, D. C. J . Mol. Biol. 1982, 160, 181-217. Tainer, J. A,; Getzoff, E. D.; Richardson, J. S.; Richardson, D. C. Nature (London) 1983, 306, 284-287. Oberley, L. W., Ed. Superoxide Dismutase; CRC Press: Boca Raton, FL, 1982. Cudd, A,; Fridovich, I. J . Biol. Chem. 1982, 257, 11443-1 1447. Getzoff, E. D.; Tainer, J. A,; Weiner, P. K.; Kollman, P. A,; Richardson, J. S.; Richardson, D. C. Nature (London) 1983, 306, 287-289. Rotilio, G.; Morpurgo, L.; Giovagnoli, C.; Calabrese, L.; Mondovi, B. Biochemistry 1972, II, 2187-2192. Rigo, A.; Viglino, P.; Rotilio, G.; Tomat, R. FEBS Lett. 1975, 50, 86-88. Calabrese, L.; Cocco, D.; Desideri, A. FEBS Lett. 1979, 106, 142-144. Strothkamp, K. G.; Lippard, S. J. Biochemistry 1981, 20, 7488-7493. Mota de Freitas, D.; Valentine, J. S . Biochemistry 1984, 23, 2079-2082. Valentine, J. S.; Mota de Freitas, D In Superoxide and Superoxide Dismutase in Chemistry, Biology, and Medicine; Rotilio, G., Ed.; Elsevier: Amsterdam, 1986; pp 149-154. But see ref 14. Valentine, J. S.; Roe, J. A.; Butler, A,; Mota de Freitas, D. In Biological and Inorganic Copper Chemistry; Kariin, K. D., Zubieta, J., Eds.; Adenine: Guiderland, NY, 1986; Vol. I, pp 53-60. But see ref 14. In ref 12 and 13, T2 measurements are interpreted as being the result of fast exchange between free and bound phosphates. From the studies reported here, it is clear that the exchange mechanism for T2is more complex and therefore that direct line shape analysis is not appropriate for this system. Bertini, I.; Luchinat, C. N M R of Paramagnetic Molecules in Biological Systems; Benjamin/Cummings: Menlo Park, CA, 1986. Malinowski, D. P.; Fridovich, I. Biochemistry 1979, 18, 5909-591 7.

0 1987 American Chemical Society

Interaction of Inorganic Phosphate with Cu,Zn-SOD

1

/5M

a

10.0

Inorganic Chemistry, Vol. 26, No. 17, 1987 2789

10.0

500

ho 1250

5.0,

0

.

0

(250 0

0

5.0

0 -3.0

I

-2.0

-1.0

--

0 0

0L -3.0

Figure 1. Dependence of TI;I and T2;l values of the 31P signal of phosphate solutions in the presence of Cu,Zn-SOD as a function of phosphate concentration at pH 8.0. The measurements were performed at 36.4 MHz. The enzyme concentration was 0.23 mM, and the phosphate concentrations varied from (4-5) X loT3to 0.4-0.7 M.

T

$

(s-ll

1

z

7

0O

-1.0

- 2.0

Figure 3. Dependence of TIC1and T2;l values of the 31P signal of phosphate solutions in the presence of Cu,Zn-SOD as a function of phosphate concentration at pH 6.3. The measurements were performed at 32.4 MHz. The enzyme concentration was 0.16 mM and the phosphate concentrations varied from (4-5) X to 0.4-0.7 M.

w T;d Is-11 ( 0 )

7

I]

I

m

'

1.0

-3.0

-2.0

- 1.0

0

.

. 150

0

m

0

Ol

-3.0

-2.0

10 0

-1.0

Figure 2. Dependence of Tl;l and T2P1values of the 31P signal of phosphate solutions in the presence of Cu,Zn-SOD as a function of phosphate concentration at pH 7.0. The experimental conditions were the same as in Figure 1.

Figure 4. Dependence of Tl;l and TZ;l values of the 31P signal of phosphate solutions in the presence of phenylglyoxal-modified Cu,ZnSOD as a function of phosphate concentration at pH 6.3. The experimental conditions were the same as in Figure 3.

relaxation times, TI, were measured with the inversion recovery method using an appropriate nonlinear least-squares fitting program. Transverse relaxation times, T2,were obtained from the line width at half-height, decreased by the line-broadening contribution resulting from exponential filtering, through the relation T2-l = TAU. (These values were confirmed by a spin-echo experiment.) pH values quoted are not corrected for the NMR spectra were measured with 4000-Hz deuterium isotope effect. 31P spectral width by using 8K data points.

lutions, are shown in Figures 1-4. The first striking feature is the much larger T2;' with respect to T1L1values at every phosphate c~ncentration.'~This result indicates that T1;I is not determined by T ~ - I , the exchange rate between bound and free phosphate. Otherwise, TI;l and T2;' would be equal and equal to ~ M T M - ' ,where fM is the molar fraction of bound phosphate. Therefore the upper limit of T ~ - 'is (fMTI,)-l,which is equal to lo2 s-I.15 Since in all cases the relaxation parameters are concentration dependent, the TIc1data can also be used to estimate the affinity of phosphate for the protein. The T1data have been used for this purpose because we believe that we understand their significance, but similar values are obtained by analyzing the T2 data. The data at pH 8 and 7 for the native protein show a simple sigmoidal behavior (Figures 1 and 2), indicating the presence of a single binding site for phosphate within the sphere of influence of the paramagnetic copper ion. The concentration dependence is satisfactorily fit by the following equation for the simple equilibrium enzyme + phosphate + enzyme-phosphate complex:

Results The binding of phosphate anion to bovine Cu,Zn-SOD was followed by the changes in the nuclear relaxation rates of the 31P nucleus of phosphate as measured by 32.4- or 36.4-MHz 31P N M R . Experiments were performed at pH 8.0, 7.0, and 6.3 in the presence of the native protein and pH 6.3 in the presence of the phenylglyoxal-modified protein. 31PTI and T2values were measured at SOD concentrations of 0 . 1 6 4 2 3 mM and phosphate to 0.4-0.7 M. In all concentrations ranging from (4-5) X instances, the TI-l and T2-I values were drastically increased with respect to analogous solutions containing the reduced proteins or no proteins at all, the latter two being almost indistinguishable. We conclude, therefore, that the increase in nuclear relaxation rates is entirely due to the interaction of phosphate with the paramagnetic copper(I1) center in the active site. The paraand T2p1, magnetic contributions toward the relaxation rates, T1cl obtained by subtracting from the experimental data on the oxidized protein solutions those of the corresponding reduced protein so-

TI,l

= cE[TIM-'K/(l

+ KC,)]

(1)

where C, and CI are the enzyme and phosphate concentrations, respectively, K is the affinity constant of phosphate for its binding site, and is the relaxation rate of bound phosphate. The best fit values are K = 20 i 4 and 34 k 3 M-' and TIM-I= 900 f 140 and 836 i 50 s-' at pH 8.0 and 7.0, respectively. Although

2790 Inorganic Chemistry, Vol. 26, No. 17, 1987 the affinity changes somewhat on passing from pH 8 to 7, TI;' changes accordingly so that the TIM-'values turn out to be the same within experimental error. The analysis of the relaxation data in terms of the possible contributing mechanisms may provide information about the and CfMTzp)-' contain a phosphate binding site. Both CfMTlp)-' dipolar contribution, which is given by the coupling between the unpaired electron located on the metal and the 3'P nucleus. Such a mechanism is well understood theoretically and can be used to infer structural information. The ratio between the dipolar contributions to CfMTZp)-' and CfMTlp)-'can be calculated if the correlation time T~ for the coupling is known. In this case, the correlation time has been independently measured on the free enzyme through water 'H nuclear magnetic resonance dispersion measurements and found to be the electronic relaxation time of copper(I1) in the compound (1.8 X lo4 s at room temperat~re).'~ If it is assumed that this time is the same in the phosphate adduct and in the free enzyme, the above ratio is 1.27. Since every further mechanism operating on TI-' or T2-' or both leads to additional contributions to the rates, it follows that Tcl, being much larger than TI-', cannot be determined solely by the dipolar relaxation mechanism. Therefore we use only the TI-' data to deduce structural information. The TIMc'values obtained from eq 1 can be related to the metal-phosphorus distance, r, through the Solomon equation (eq s (see above), po is the 2). In this equation, T , is 1.8 X TIM-'=

permeability of vacuum, r is the distance between the 31Patom of phosphate and the copper(I1) ion, yN is the nuclear magnetogyric ratio, g, is the electron g factor, pB is the Bohr magneton, S is the spin quantum number, and oIand usare the nuclear and electronic Larmor frequencies. It has already been shown that this equation is valid for copper(I1) systems at the magnetic fields of the present experiments (1.88-2.1 1 T).I7 The estimated distance is 5.3 8,. Such a value indicates that phosphate is not directly bound to the copper ion and confirms a posteriori the assumptions of purely metal-centered dipolar interaction (Le. no spin delocalization contributions through chemical bonds) and of the constancy of the electronic relaxation time, since the copper chromophore is not directly perturbed, as already suggested from electronic and EPR spectroscopies.'' Furthermore, the 'H N M R spectra of Cu,Co-SOD (a derivative of Cu,Zn-SOD in which Zn(I1) has been replaced by Co(I1)) measured in the presence of excess phosphate are absolutely identical with the spectra in the absence of phosphate. Such spectra display the signals of all of the protons of the histidyl imidazole ligands coordinated to both cobalt and copper and thus are highly sensitive to changes in coordination of the metal centers.'* This experiment shows that phosphate does not alter the coordination sites. It should also be noted that a similar study of the interaction of phosphate with copper(I1)-substituted carbonic anhydrase under conditions comparable to those described here for Cu,Zn2SOD has been reported." In that case, phosphate is believed to bind directly to copper and, accordingly, CfMTI,)-lvalues were observed that were 20 times larger than those reported here for Cu2ZnzSOD. The TI;' data at p H 6.3 (Figure 3) show a biphasic behavior that cannot be fit by eq 1. The pattern is suggestive of the presence of two anion binding sites with different affinities. Assuming that the two sites are not interacting, i.e. that binding of phosphate to one site does not significantly alter the affinity of the other site for phosphate nor its interaction with the paramagnetic center, (17) Bertini, I.; Briganti, F.; Luchinat, C.; Mancini, M.; Spina, G. J . Magn.

Reson. 1985, 63, 41-55. (18) Bertini, I.; Lanini, G.; Luchinat, C.; Messori, L.; Monnanni, R.; Scozzafava, A. J . A m . Chem. SOC.1985, 107, 4391-4397. (19) Bertini, I.; Luchinat, C.; Scozzafava, A. FEBS Lett. 1978, 93, 251-256.

Mota de Freitas et ai. the data indicate a site at a distance greater than 7 8, from the metal with an affinity constant greater than 100 M-' and another at -S 8, with an affinity of 10 M-I. The latter site is likely to be the same as that characterized at higher pH values. Analogous behavior is shown by phosphate interacting with the phenylglyoxal-modified protein at the same pH (Figure 4). The curve is lower, however, and shifted to higher phosphate concentration, indicating that the 31PN M R data are probably monitoring the same sites as in the native enzyme and that the affinity for both of them is reduced by about a factor of 3. A more quantitative analysis of the pH 6.3 data would be difficult to attempt and is probably not worthwhile in view of the fact that the more biochemically relevant binding site is likely to be the one closer to the catalytic metal and that the binding behavior has been satisfactorily characterized at physiological pH. The T2;' data are also shown in Figures 1-4. As noted above, they are 70 times higher than the corresponding TI;' values. Such a large difference is difficult to explain. In principle, it may be due to exchange broadening operating on T2 according to

where AoM is the difference in chemical shift between free and protein-bound phosphate. Exchange broadening is maximum when 7M-l = AuM. In such a case, the experimental T2;' values are ~ = lo5 rad s-' ( N SO0 ppm reproduced by eq 3 for T ~ =- AwM at 32 MHz). This is a lower limit for AwM. Such a value is unrealistically high for phosphate that is not directly coordinated to the copper center. In addition, the phosphate signal would have shown an experimental shift of at least 100 Hz, whereas no shift could be detected in the present case. The temperature and field dependence of the line width should help in clarifying the nature of the T2-' mechanism. However, the line width turns out to be essentially insensitive to temperature in the range 4-27 O C and to the magnetic field between 2.1 1 and 7.05 T. It is surprising that the concentration dependence of T2;' always parallels, at least qualitatively, that of TI;', suggesting that the line broadening is in any case related to the same kind of sites giving rise to detection of paramagnetic effects on TI;'. There are other 31PN M R studies in which it has been noted that T2p-' is much larger than TI;' for a phosphate group that is not coordinated to a paramagnetic metal ion.20 The reasons for such a difference have not been analyzed. Clearly, further efforts should be made to understand this phenomenon, but they are outside the scope of the present work. Whatever the nature of the line broadening is, the conclusions drawn from the TI;] data should be absolutely reliable for obtaining structural inf~rmation.'~ Discussion The interaction of phosphate with the free amino acid arginine has been demonstrated by observation of small chemical shifts of the 3'P resonance near the phosphate pK,, region. This effect was attributed to preferential binding of the dibasic phosphate anion by the guanidinium group of In the present study, the interaction of phosphate with bovine Cu,Zn-SOD was detected by means of the increase in the nuclear relaxation rates of the 31Pnucleus due to the proximity of the paramagnetic copper(I1) center.I5 Using the measured TI;' values, we have determined that, at pH 7.0 and 8.0, phosphate binds at a specific site, which is located approximately 5 8, from the copper ion. It is known from the X-ray crystallographic studies3 that the guanidinium group of Arg-141 is located approximately 5 8, from the copper(I1) ion. Previous studies have also demonstrated that modification of Arg- 141 with phenylglyoxal diminishes the influence of phosphate on the SOD activity and the anion-binding properties of the protein." These results combined with our (20) Bean, B. L.; Koren, R.; Mildvan, A. S. Biochemistry 1977, 16, 3323-3329. (21) Katz, R.; Yeh, H. J. C.; Johnson, D. F. Biochem. Biophys. Res. Commun. 1974, 58, 316-321. (22) Katz, R.; Yeh, H. J. C.; Johnson, D. F. Mol. Pharm. 1977,13,615-620.

Interaction of Inorganic Phosphate with Cu,Zn-SOD observation that this modification diminishes the affinity constant for phosphate binding suggest strongly that Arg-141 is the site of phosphate binding. The concentration dependences of TI;) parameters for the native protein solutions (Figures 1 and 2) at pH 8.0 and 7.0 yield phosphate affinity constants of 20 f 4 and 34 f 3 M-I, respectively. The calculated phosphate-binding constants are thus significantly higher than those previously obtained for the interaction between phosphate and free arginine, Le., K = 5 M-'.21 This observation suggests that Arg-141 is activated toward phosphate binding, probably as a result of its location in a hydrophobic region of the solvent channel in the protein. At pH 6.3, two binding constants for phosphate are detected. If we consider that the lower affinity value at pH 6.3 corresponds to the single values observed at pH 7.0 and 8.0, we must conclude that the affinity of this site for phosphate reaches a maximum between pH 6.3 and 8.0. The increase in affinity for phosphate between pH 6.3 and 7.0 is probably due to a higher affinity for HP042- relative to that for H2P04-. The decrease in affinity between pH 7.0 and 8.0 is more difficult to explain. It has been reported that the SOD activity of the protein decreases above pH 7.23 A possible explanation for both effects is deprotonation of a group other than Arg-141 near the active-site channel. The nature of the second binding site for phosphate at pH 6.3 cannot be identified from the present N M R study. It is interesting to compare our conclusions with those obtained from studies of phosphate binding to Co,Zn-SOD, a derivative in which Co2+has replaced Cuz+ at the native copper site. In this derivative, it has been concluded that phosphate interacts directly with the cobalt chromophore, on the basis of e l e c t r o n i ~ ,ESR:4 ~~*~~ and 31PNMR25,26spectroscopic evidence. On the basis of this evidence, it has been suggested that phosphate binds directly to the cobalt ion and that this direct binding induces the rupture of the imidazolate bridge between cobalt and zinc. These results (23) Mug, D.; Rabani, J.; Fridovich, I. J. Biol. Chem. 1972, 247,48394842. (24) Desideri, A.; Cocco, D.; Calabrese, L.; Rotilio, G. Biochim. Biophys. Acta 1984, 785, 1 1 1-1 17. (25) Hirose, J.; Hayakawa, C.;Noji, M.; Kidani, Y . Inorg. Chim. Acta 1985, 107, L7-L10. (26) Banci, L.; Bertini, I.; Luchinat, C.; Monnanni,R.; Scozzafava,A. Inorg. Chem. 1987, 26, 153-156.

Inorganic Chemistry, Vol. 26, No. 17, 1987 2791 are thus in sharp contrast to those obtained for the native protein, where phosphate does not appear to bind to copper but is 5 A away at the guanidinium group of Arg-141. These results are, of course, consistent with the observation of pronounced electronic spectral changes in the visible region upon binding of phosphate to Co,Zn-SOD whereas no visible spectral changes are observed upon binding of phosphate to Cu,Zn-SOD.I' Quantitative analysis of phosphate binding to the argininemodified protein provides evidence that chemical modification of Arg-141 with phenylglyoxal is lowering its affinity for phosphate. The copper-phosphate distance is essentially the same as in the unmodified protein. Chemical modification of an arginine residue with phenylglyoxalI6 places steric hindrance around the arginine side chain without altering its positive charge. We therefore believe that phosphate may still be attracted by the positive charge of the modified arginine residue but that the chemical modification prevents a specific molecular interaction between the side chain and phosphate, resulting in a lower phosphate affinity and, thus, smaller Tl;l values for the arginine-modified protein. Smaller relaxation values cannot be due to the presence of some unmodified native protein present as a contaminant in the preparation of the arginine-modified SOD since the TI;' profile would only be lowered and not shifted, as is observed. The functional significance of the binding of phosphate to bovine Cu,Zn-SOD is not immediately apparent to us. Although it is clear from our earlier work" that phosphate at concentrations typically present in the cell will cause partial inhibition of the SOD activity of this protein, it is also possible that phosphate may play a role in some as yet unknown physiological function of this protein. Future studies hopefully will clarify the role of phosphate in the function of this protein. Acknowledgment. We are grateful to Dr. Keiko Kanamori for helpful discussions during the progress of this project. The technical assistance of Mr. Nazim Jaffer is also highly appreciated. This work was supported by Grant GM-28222 from the National Institutes of Health and by a NATO travel grant. Fellowship support for D.M.d.F. from the Exxon Education Fund, Calouste Gulbenkian Foundation, and INIC/Invotan (Portugal) is gratefully acknowledged. Registry No. Cu,Zn-SOD, 9054-89-1; Arg, 74-79-3; phosphate, 14265-44-2.